[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-2503":3},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"view_count":33,"doi":34,"paper":35,"created_at":53},2503,"Agentic AI for Livestock Housing Management: Applications, Benchmarking, and Readiness Assessment","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112361","Agentic artificial intelligence is emerging as an extension of Precision Livestock Farming by linking perception, reasoning, planning, and bounded action within human-supervised livestock-housing workflows. This review synthesizes 90 publications on agentic AI, multi-agent systems, retrieval-augmented generation, large language models, foundation models, robotics, digital twins, simulation, computer vision, cyber-physical control, and related enabling technologies for livestock-housing management. We propose a Perception–Reasoning–Action–Safety (PRAS) loop and an Agentic Livestock Housing Readiness Scale to classify systems from passive monitoring and advisory decision support to supervised, safety-constrained closed-loop operation. A staged benchmarking perspective is also used to integrate algorithmic performance, biological relevance, safety, auditability, economic feasibility, and human–AI interaction. Current evidence is strongest for perception, advisory reasoning, natural-language data access, welfare-risk interpretation, and simulation-supported decision support, whereas robust barn-wide autonomous control remains largely unvalidated. Technology categories were coded non-exclusively; therefore, publication frequencies indicate representation within the selected corpus rather than effectiveness, evidence strength, or deployment readiness. Across species, dairy cattle provide the most developed evidence base, poultry studies mainly address environmental comfort and nutrition support, swine systems emphasize simulation-based precision feeding, and small-ruminant evidence remains concentrated in advisory tools and contextual embodied monitoring. Overall, agentic AI in livestock housing is currently more mature as an orchestration, explanation, and decision-support layer than as an autonomous control technology. By distinguishing direct housing applications, semi-agentic prototypes, and enabling technologies, this review clarifies the gap between current evidence and deployable autonomy. Progress towards higher readiness will require cross-farm validation, biological plausibility, source-grounding audits, safety assurance, interoperability, economic assessment, transparent benchmarking, and explicit human oversight.","智能体人工智能（Agentic AI）正在成为精准畜牧养殖（Precision Livestock Farming）的延伸，通过在人工监督的畜舍工作流程中将感知、推理、规划与有限行动相连接。本综述综合了90篇文献，涵盖智能体人工智能、多智能体系统、检索增强生成、大语言模型、基础模型、机器人技术、数字孪生、仿真、计算机视觉、信息物理控制及相关使能技术在畜舍管理中的应用。我们提出了感知—推理—行动—安全（Perception–Reasoning–Action–Safety, PRAS）循环和智能体畜舍就绪度量表（Agentic Livestock Housing Readiness Scale），将系统从被动监测与建议性决策支持分类至受监督、安全约束的闭环运行。同时采用分阶段基准测试视角，整合算法性能、生物学相关性、安全性、可审计性、经济可行性及人机交互。当前证据最充分的领域为感知、建议性推理、自然语言数据访问、福利风险解读及仿真支持的决策支持，而稳健的畜舍级自主控制仍基本未经验证。技术类别采用非排他性编码，因此文献频次仅表示在所选语料中的代表性，而非有效性、证据强度或部署就绪度。跨物种来看，奶牛提供了最成熟的证据基础，家禽研究主要涉及环境舒适度与营养支持，猪系统侧重于基于仿真的精准饲喂，而小反刍动物的证据仍集中于建议性工具与情境化具身监测。总体而言，畜舍中的智能体人工智能目前作为编排、解释与决策支持层比作为自主控制技术更为成熟。通过区分直接畜舍应用、半智能体原型与使能技术，本综述厘清了当前证据与可部署自主性之间的差距。迈向更高就绪度需要跨农场验证、生物学合理性、来源溯源审计、安全保障、互操作性、经济评估、透明基准测试及明确的人工监督。",null,"Computers and Electronics in Agriculture","2026-09-14T00:00:00Z","论文",10,false,86,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},22,23,18,14,9,1,"该综述系统梳理了智能体AI在畜禽舍管理中的应用，提出PRAS框架与准备度评估量表，方法新颖、结论审慎，对智慧畜牧研究与落地具有较高参考价值。",[25],{"name":10,"url":6},[27,28,29,30,31,32],"智慧农业","农业人工智能","多智能体","数字孪生","智能畜牧","精准养殖",0,"10.1016\u002Fj.compag.2026.112361",{"doi":34,"openalex_id":36,"authors":37,"venue":10,"cited_by_count":33,"oa_url":44,"card":45,"direction":51,"ingested_from":52},"W7213126233",[38,40,42],{"name":39,"orcid":9},"Alexey Ruchay",{"name":41,"orcid":9},"Hao Guo",{"name":43,"orcid":9},"Andrea Pezzuolo","https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fpii\u002FS0168169926009592\u002Fpdf",{"tldr":46,"method":47,"finding":48,"direction":49,"opportunity":50},"综述90篇文献，提出PRAS框架与就绪度量表，评估智能体AI在畜舍管理中的应用与成熟度。","文献综述，提出感知-推理-行动-安全循环与就绪度量表，分阶段基准评估。","智能体AI目前更适合作为决策支持与解释层，而非自主控制，牛的证据最充分。","农业人工智能与决策模型","可研究跨农场验证、安全约束闭环控制与可审计基准，填补自主畜舍管理空白。","智慧农业 \u002F 农业物联网","openalex","2026-09-15T23:30:01.732357Z"]